On 2/3/2020 10:21 AM, Ferruh Yigit wrote:
> On 2/2/2020 8:23 AM, Ori Kam wrote:
>> Hi Ferruh,
>>
>> PSB,
>> Thanks,
>> Ori
>>
>>> -----Original Message-----
>>> From: Ferruh Yigit <ferruh.yi...@intel.com>
>>> Sent: Friday, January 31, 2020 1:08 PM
>>> To: Ori Kam <or...@mellanox.com>; Wenzhuo Lu <wenzhuo...@intel.com>;
>>> Jingjing Wu <jingjing...@intel.com>; Bernard Iremonger
>>> <bernard.iremon...@intel.com>
>>> Cc: dev@dpdk.org; Slava Ovsiienko <viachesl...@mellanox.com>;
>>> sta...@dpdk.org
>>> Subject: Re: [PATCH] app/testpmd: fix uninitialized members of mpls
>>>
>>> On 1/30/2020 4:59 PM, Ori Kam wrote:
>>>> Some of the memebers of the mpls struct are not initialized.
>>>> this commit init the uninitialized members.
>>>>
>>>> Coverity issue: 325735
>>>> Fixes: 3e77031be855 ("app/testpmd: add MPLSoGRE encapsulation")
>>>> Cc: sta...@dpdk.org
>>>>
>>>> Signed-off-by: Ori Kam <or...@mellanox.com>
>>>> ---
>>>> app/test-pmd/cmdline_flow.c | 4 +++-
>>>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/app/test-pmd/cmdline_flow.c b/app/test-pmd/cmdline_flow.c
>>>> index e99e24c..c2cc4c5 100644
>>>> --- a/app/test-pmd/cmdline_flow.c
>>>> +++ b/app/test-pmd/cmdline_flow.c
>>>> @@ -4576,7 +4576,9 @@ static int comp_set_raw_index(struct context *,
>>> const struct token *,
>>>> struct rte_flow_item_gre gre = {
>>>> .protocol = rte_cpu_to_be_16(ETHER_TYPE_MPLS_UNICAST),
>>>> };
>>>> - struct rte_flow_item_mpls mpls;
>>>> + struct rte_flow_item_mpls mpls = {
>>>> + .ttl = 0,
>>>> + };
>>>
>>> why not use "= {0};", the end result will be same, struct will be all
>>> zeroed out
>>> in both case, "= {0};" makes the intention more obvious I think.
>>>
>>
>> On some compilers this kind of initialization result in an error:
>> error: missing braces around initializer
>
> As far as I remember, that is triggered with old compiler, when you are
> initializing array of structs, or first element of the struct is an array
> (which
> is the case for mpls struct), in that case ={{0}} was solving the issue.
>
> Anyway, I got the concern, the patch looks good to me.
Also seems "={}" is working in that case, thanks to Stephen, it is cleaner than
"={{0}}", what do you think about using "={}" ?
>
>> it looks like a known issue of GCC
>> So I just prefer to init the relevant fields.
>>
>>>> uint8_t *header;
>>>> int ret;
>>>>
>>>>
>>
>
